PfCRT and the trans-vacuolar proton electrochemical gradient: regulating the access of chloroquine to ferriprotoporphyrin IX

Summary

It is accepted that resistance of Plasmodium falciparum to chloroquine (CQ) is caused primarily by mutations in the pfcrt gene. However, a consensus has not yet been reached on the mechanism by which resistance is achieved. CQ-resistant (CQR) parasite lines accumulate less CQ than do CQ-sensitive (CQS) parasites. The CQR phenotype is complex with a component of reduced energy-dependent CQ uptake and an additional component that resembles energy-dependent CQ efflux. Here we show that the required energy input is in the form of the proton electrochemical gradient across the digestive vacuole (DV) membrane. Collapsing the DV proton gradient (or starving the parasites of glucose) results in similar levels of CQ accumulation in CQS and CQR lines. Under these conditions the accumulation of CQ is stimulated in CQR parasite lines but is reduced in CQS lines. Energy deprivation has no effect on the rate of CQ efflux from CQR lines implying that mutant PfCRT does not function as an efflux pump or active carrier. Using pfcrt-modified parasite lines we show that the entire CQ susceptibility phenotype is switched by the single K76T amino acid change in PfCRT. The efflux of CQ in CQR lines is not directly coupled to the energy supply, consistent with a model in which mutant PfCRT functions as a gated channel or pore, allowing charged CQ species to leak out of the DV.

Introduction

Plasmodium falciparum malaria currently accounts for an estimated one to three million deaths each year, most of these being African children. The spread of P. falciparum strains resistant to chloroquine (CQ) has significantly contributed to these distressing mortality figures (Trape, 2001). Despite the devastating impact of CQ resistance, the underlying molecular mechanism remains controversial.

It is now accepted that the membrane protein, PfCRT (P. falciparum chloroquine resistance transporter) is the primary determinant of CQ resistance (Fidock et al., 2000; Sidhu et al., 2002; Lakshmanan et al., 2005). PfCRT is located on the membrane of the digestive vacuole (DV) of the intracellular parasite (Fidock et al., 2000; Cooper et al., 2002). Inside the DV, CQ binds to ferriprotoporphyrin IX (FP) that is released during haemoglobin proteolysis (Bray et al., 1999). Free FP is toxic, causing lipid peroxidation and membrane damage. The parasite counters these toxic effects by converting FP into non-toxic haemozoin biocrystals (Slater and Cerami, 1992; Dorn et al., 1998; Sullivan et al., 1998; Pagola et al., 2000). The binding of CQ to FP inhibits this biocrystallization process and the ensuing build up of FP and CQ–FP complexes ultimately leads to the death of the parasite (Slater and Cerami, 1992). In CQ-resistant (CQR) parasite lines, mutations within PfCRT, in particular a conserved K76T (lysine to threonine) amino acid substitution in a predicted transmembrane region appear to reduce the binding of CQ to its FP target, thereby protecting the parasite from the toxic effects of the drug (Bray et al., 1998; Lakshmanan et al., 2005).

This critical PfCRT K76T mutation is predicted to result in the loss of a positive charge in the transmembrane region of a putative pore. It was suggested that the K76T charge loss mutation would allow CQ to passively leak down a concentration gradient, away from the FP target (Warhurst et al., 2002). This would reduce binding of CQ to FP, rendering the parasite resistant to the drug. A recent study underlined the importance of charged residues in PfCRT transmembrane regions: parasite lines that were previously CQR became fully CQ-sensitive (CQS) with restored CQ–FP binding when selected for resistance to halofantrine (HF) or amantadine. These lines still contained the K76T mutation but were found to harbour an additional mutation (S163R) that reintroduced a positive charge in a transmembrane region of the protein (Johnson et al., 2004).

Bioinformatics studies suggest that PfCRT might be more similar to members of the drug-metabolite transporter (DMT) superfamily than to a channel (Martin and Kirk, 2004; Tran and Saier, 2004). The idea of direct CQ transport through PfCRT is compatible with the charge substitution findings outlined above and is also consistent with recent trans-stimulation data suggesting that CQ resistance might be a carrier-mediated process (Sanchez et al., 2003; 2004; 2005). These data, together with evidence that the CQ resistance mechanism requires an input of energy (Bray et al., 1992; Krogstad et al., 1992; Sanchez et al., 2003), have led some authors to propose that mutant PfCRT is a CQ efflux pump or active carrier, acting against a passive inward leak of uncharged drug to actively transport CQ out of the resistant parasite (Sanchez et al., 2003; 2005; Naude et al., 2005).

In this study, we have examined the energy coupling and subcellular localization of the CQ uptake process and the CQ resistance mechanism. Our results indicate that the passive exit of CQ through modified PfCRT lowers the vacuolar drug concentration and reduces the binding of CQ to FP. Surprisingly the outward transport of CQ in resistant parasite lines does not appear to be directly coupled to the energy supply. Therefore our results are inconsistent with the efflux pump or active carrier models and can more readily be explained by a model in which mutated PfCRT is a gated aqueous pore that permits the passive outward movement of protonated forms of the drug.

Results

The effect of uncoupling the proton motive force on CQ accumulation in CQS and CQR parasite lines

Data presented in Fig. 1 show the effect of glucose starvation on the uptake of CQ by CQS and CQR parasite lines. Figure 1A shows data for the non-transformed Dd2 (CQR) and GC03 (CQS) lines. Glucose starvation reduces the accumulation of CQ in the GC03 CQS parasites but stimulates CQ accumulation in the Dd2 CQR parasites. These data are consistent with previously published reports on the effect of glucose starvation in CQS and CQR lines (Bray et al., 1992; 1996; Krogstad et al., 1992; Sanchez et al., 2003; 2004). Figure 1B shows that the effects of glucose starvation in the CQS line are changed to resemble those of a CQR line if the wild-type pfcrt allele is exchanged for the South-east Asian Dd2 CQR pfcrt allele. Data presented in Fig. 1C show that reversing the K76T mutation is all that is required to change the energy dependence of CQ uptake in the CQR line so that it resembles that of a CQS line. These data show that the amino acid change at codon 76 of the pfcrt gene plays a key role in the utilization of energy to control CQ accumulation, just as it does in the rest of the CQ susceptibility phenotype (Lakshmanan et al., 2005).

Fig. 1.

The effect of glucose starvation on the accumulation of [³H]-CQ by CQS and CQR lines of P. falciparum. Parasite lines were incubated with or without 10 mM glucose and CQ accumulation was measured over a 10 min period using a concentration of 2 nM [³H]-CQ. The CQ cellular accumulation ratio (CAR) is the ratio of the intracellular versus the extracellular [³H]-CQ concentration, [CQ]_in/[CQ]_out, over a 10 min period.

A. Data for the non-transformed Dd2 (CQR) and GC03 (CQS) lines.

B. Data for the C2^GC03 and C3^Dd2 lines (produced by allelic modification of the pfcrt locus in the GC03 line) that express the GC03 wild-type CQS and the Dd2 mutant CQR pfcrt alleles respectively (Sidhu et al., 2002).

C. Data for the C-1^Dd2 and T76K-1^Dd2 recombinant clones (produced by allelic modification of the pfcrt locus in Dd2 parasites), with C-1^Dd2 expressing the parental Dd2 allele whereas T76K-1^Dd2 expresses a variant Dd2 ‘back-mutant’ allele in which the mutant threonine codon 76 (T76) was replaced with the wild-type lysine codon 76 (K76) (Lakshmanan et al., 2005).

Histogram bars represent means ± standard deviation of at least five independent experiments, each performed in triplicate. The effect of glucose starvation was significant in all cases (Mann–Whitney, P < 0.001).

Martin and Kirk (2004) suggested that PfCRT most closely resembles certain members of the DMT superfamily. Transporters belonging to this family are not thought to be primary active transporters (energized directly by ATP), but many of them are known to be secondary active transporters, being energized by the proton electrochemical gradient (Jack et al., 2001). For these reasons we have examined the influence of the proton electrochemical gradient on the accumulation of CQ in CQS and CQR lines. Data presented in Fig. 2A–C show the effect of 10 μM carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP) on the uptake of CQ by the same sets of CQS and CQR lines. This agent closely mimics the effect of glucose starvation in all the parasite lines tested. FCCP is a potent proton uncoupler (Benz and McLaughlin, 1983) that is commonly used at micromolar concentrations to dissipate proton gradients of intracellular organelles and that has been used for over 20 years to dissipate lysosomal proton gradients (Yamashiro et al., 1983). We cannot, however, rule out secondary effects of collapsing the DV proton gradient on other gradients such as calcium or potassium. Neither can we rule out a direct effect of FCCP on PfCRT itself. Nonetheless, these data strongly suggest that the DV proton electrochemical gradient plays a key role in the uptake of CQ in CQS parasites and in the efflux of CQ from CQR parasites.

Fig. 2.

The effect of FCCP on the accumulation of [³H]-CQ by CQS and CQR lines of P. falciparum. Parasite lines were incubated with 10 mM glucose, with or without 10 μM FCCP and CQ accumulation was measured over a 10 min period using a concentration of 2 nM [³H]-CQ. For all experiments, CQ cellular accumulation ratio (CAR) is given as the ratio of the intracellular versus the extracellular [³H]-CQ concentration, [CQ]_in/[CQ]_out, over a 10 min period.

A. Data for the non-transformed Dd2 and GC03 lines.

B. Data for the C2^GC03 and C3^Dd2 lines.

C. Data for the C-1^Dd2 and T76K-1^Dd2 recombinant clones.

Histogram bars represent means ± standard deviation of at least seven independent experiments, each performed in triplicate. The effect of FCCP was significant in all cases (Mann–Whitney, P < 0.001).

Compartmentalization of CQ accumulation in CQS and CQR parasite lines

It is widely assumed that the vast majority of CQ that is accumulated by P. falciparum-infected erythrocytes is located in the DV but this has proven difficult to convincingly demonstrate. Sullivan et al. (1996) showed that [³H]-CQ binds to haemozoin crystals in the DV and that a high CQ concentration is reached after 24 h of continuous incubation in culture. However, little is known about the initial intracellular distribution of [³H]-CQ during transport experiments that are typically performed over several minutes. In the only study of its kind, Saliba et al. (1998) showed that isolated DV are capable of accumulating CQ when energized with ATP. However, the level of CQ accumulation seen in these isolated DV was only about one-twentieth of that seen with intact infected erythrocytes (Saliba et al., 1998). It may be that the DV is not the major site of CQ accumulation after all. Alternatively, as stated by the authors the isolation procedure may have interfered with a process that is important for CQ accumulation (Saliba et al., 1998). Also, although they contain a full complement of proteolytic enzymes (Goldberg et al., 1991), isolated DV have been cut off from the supply of haemoglobin. It is likely that any residual haemoglobin will have been rapidly metabolized soon after isolation and that any FP that was released will probably be in an aggregated or crystalline form. Indeed, Saliba et al. were unable to detect any haemoglobin in their isolated vacuoles (Saliba et al., 1998). Thus there was probably very little free FP available for CQ binding, potentially explaining the low levels of CQ uptake seen in their study. For these reasons we have developed a rapid protocol in which parasites are freed from their host cells and DV isolated from the free parasites after loading the infected cells with [³H]-CQ. Leakage of pre-loaded [³H]-CQ was minimized by optimizing the conditions for speed rather than purity and by keeping the temperature at 0–4°C during the fractionation steps.

The results of a representative set of these experiments are presented in Fig. 3. Data are presented for the CQS C2^GC03 line in Fig. 3A. This recombinant clone was generated from the CQS GC03 line that was transfected with its own wild-type pfcrt allele and that displays the full CQS phenotype (Sidhu et al., 2002). In this CQS line, most (approximately 85%) of the [³H]-CQ taken up by the parasite during the 10 min pre-incubation period is located in the isolated DV. Furthermore, the proton uncoupler FCCP appears to inhibit CQ accumulation primarily in the DV compartment. The slight reduction of CQ uptake seen with verapamil (VP) also appears to be primarily in the DV compartment but in this case the effects are not significant. These data finally establish the DV as the major compartment for the accumulation of CQ in CQS isolates and show directly that uncoupling the proton motive force can significantly reduce the accumulation in this compartment. Similar results were found with the other CQS parasite lines GC03 and 3D7 (data not shown).

Fig. 3.

The DV is the major CQ accumulating compartment in P. falciparum. (A and B) Data for the C2^GC03 and C3^Dd2 lines. In these experiments, parasite suspensions were incubated with 2 nM [³H]-CQ, with or without 10 μM VP or 10 μM FCCP for a 10 min period prior to subcellular fractionation. Histogram bars represent means ± standard deviation of at least eight independent experiments, each performed in duplicate. The effect of FCCP was significant in all cases (Mann–Whitney, P < 0.001). The effect of VP was significant only in the CQR line (Mann–Whitney, P < 0.001).

Data are presented for the CQR C3^Dd2 line in Fig. 3B. This recombinant clone was also generated from the GC03 line but in this case was transfected with the Dd2 CQR pfcrt allele and acquired a full CQR phenotype (Sidhu et al., 2002). Several differences are immediately apparent when compared with the data in Fig. 3A. In this CQR line, only half (53%) of the CQ in the parasite is located in the DV compartment. While the cytoplasmic CQ concentration appears to be similar in the two lines, the concentration in the DV is almost sixfold lower in the CQR line. FCCP and VP appear to exert their effects primarily in the DV compartment but this time CQ accumulation is clearly stimulated rather than inhibited.

Trans-stimulation phenomena and the direction of CQ transport

Support for active transport of CQ through PfCRT in CQR parasite lines has come from the demonstration of trans-stimulation of [³H]-CQ uptake after the parasites were pre-loaded with high concentrations of unlabelled CQ (Sanchez et al., 2003; 2004; 2005). We also observe a trans-stimulation of [³H]-CQ uptake after pre-loading with unlabelled CQ. A representative graph is shown in Fig. 4A. These data show the occurrence of trans-stimulation of [³H]-CQ uptake in the Dd2 CQR strain over a period of 4 min after pre-loading the cells with unlabelled CQ. Similar to the report of Sanchez et al. (2005), we find a trans-stimulation phenomenon in all the other CQR field isolates and in transfected parasite lines bearing CQR alleles of pfcrt (data not shown). Also like that report, we did not observe this phenomenon in CQS field isolates or in transfected parasites harbouring wild-type pfcrt alleles (data not shown).

Fig. 4.

A pre-loading step is required for the trans-stimulation effect.

A. Data for the C3^Dd2 line. Accumulation of [³H]-CQ (1 nM) was measured after pre-loading with unlabelled CQ and washing as described (Sanchez et al., 2003).

B. Data for the same C3^Dd2 line, this time with the pre-loading step omitted and the labelled and unlabelled CQ mixed prior to incubation with parasites.

C. Data for various transformed and non-transformed CQR lines, obtained using the same protocol as in (B).

For all experiments, [³H]-CQ accumulation was measured after 4 min of incubation (given as the ratio of the intracellular versus the extracellular [³H]-CQ concentration, [CQ]_in/[CQ]_out, at 4 min). Accumulation of label was plotted as a function of the pre-equilibrium unlabelled CQ concentration. Uptake studies were performed in medium containing 10 mM glucose. Each data point represents the mean ± standard deviation of duplicate observations from at least four independent experiments.

However, we did not observe trans-stimulation of CQ uptake over a 4 min period if the [³H]-CQ and the unlabelled CQ were mixed together before the incubation (Fig. 4B). These data are representative of those observed in other CQR field isolates and transfected lines (Fig. 4C) and they strongly suggest that the unlabelled CQ and the [³H]-CQ must be on opposite sides of the membrane for the trans-stimulation effect to take place.

Efflux of CQ in CQS and CQR parasites

Previously, we reported rapid CQ efflux (t_1/2 of a few minutes) in CQS and CQR strains of P. falciparum (Bray et al., 1992). This is in contrast to the study by Krogstad et al. (1987) who reported rapid CQ efflux only in CQR strains, which was 40- to 50-fold lower in CQS strains (Krogstad et al., 1987). We were able to repeat the observations of Krogstad et al. provided that the medium contained no bicarbonate and more importantly, that the cell pellet had been washed after loading and before efflux rate measurements. Previously, we showed that CQ binds to the FP receptor with a much higher apparent affinity in CQS strains compared with CQR strains (Bray et al., 1998) and hypothesized that the factor(s) responsible for these differences might also account for the more rapid rate of efflux from CQR strains reported by Krogstad et al. (1987).

We believe that freely exchangeable drug is rapidly lost from both CQS and CQR lines during the washing step. This is consistent with the observations of Sanchez et al. (2003) who reported that approximately 40% of the pre-loaded CQ was lost from CQS parasites and 50% of pre-loaded CQ was lost from CQR parasites during a similar washing procedure. After the washing step the remaining CQ is tightly bound to FP in CQS parasites. We hypothesize that the PfCRT mutations responsible for CQ resistance reduce the apparent affinity of this interaction, by allowing protonated CQ to leak out of the DV, leading to the increased CQ efflux rates observed in CQR isolates. We have tested this hypothesis directly by measuring CQ efflux from CQS isolates in the presence of 200 nM HF, a drug known to bind strongly to FP (Blauer, 1988). HF accelerates the rate of CQ efflux in the GC03 CQS strain at the same rapid rate that is seen in resistant strains (Fig. 5A). This might be achieved because HF is sufficiently lipophilic to alkalinize the DV by binding protons and shuttling them out of the acid DV (Ginsburg et al., 1989). However, we cannot rule out the alternative possibility that these CQS parasites contain an exchange transporter that drives CQ efflux in exchange for HF influx. In any case, the effect of HF in displacing CQ from its FP receptor allows almost all of the CQ to leak out of the parasites (Fig. 5A). In contrast, neither VP nor primaquine (PQ) accelerate CQ efflux, even when used at much higher concentrations than HF (Fig. 5A). These drugs are much less lipophilic than HF and probably do not alkalinize the DV, neither do they bind to FP and displace CQ. Therefore our observations suggest that when [³H]-CQ is freely exchangeable, the rate of efflux is very rapid, even in CQS strains, providing a potential explanation for our earlier observations: the rapid CQ efflux we observed in CQS strains without the initial washing step (Bray et al., 1992) probably reflects a freely exchangeable drug compartment that is removed during the washing step of Krogstad et al. (1987). This would also explain the 40- to 50-fold differences in CQ concentration between the CQS and CQR lines at the start of the efflux experiments (Krogstad et al., 1987). This is much greater than the fivefold differences in steady-state CQ accumulation that are commonly observed (Sanchez et al., 1997; Bray et al., 1998) and probably reflects the continuing loss of CQ from the pellet of CQR isolates during the washing step.

Fig. 5.

Efflux of CQ, effect of other haem-binding drugs and effect of energy starvation.

A. Data for the CQS GC03 line measured in the presence or absence of 200 nM HF or 10 μM PQ or 10 μM VP.

B. Data for the Dd2 CQR line measured in the presence or absence of 10 mM glucose or 10 mM glucose with or without 10 μM FCCP or 10 mM glucose with or without 5 μM VP.

C. Data for the GC03 CQS line measured in the presence or absence of 10 mM glucose or 10 mM glucose with or without 10 μM FCCP. Data points are means of duplicate observations from at least two independent experiments.

Nonetheless, the Krogstad et al. (1987) protocol does consistently reveal the largest differences between the rate of efflux of CQ from CQS and CQR lines. For this reason, we employed their protocol to investigate the energy coupling of the CQ efflux process in CQS and CQR strains. In general when drug-resistant cells have evolved an active drug efflux mechanism, energy starvation slows down the rate of drug efflux so that it resembles the default drug efflux rate that is seen in drug-sensitive cells. However, in CQR P. falciparum there is no effect of withdrawing glucose or adding 10 μM FCCP on the rate of CQ efflux in the CQR (Dd2) strain (Fig. 5B). This is in contrast to the effect of 5 μM VP, which markedly slowed the rate of efflux (Fig. 5B). These data suggest that VP promotes CQ uptake by slowing the rate of CQ efflux and that a separate mechanism accounts for the stimulation of CQ uptake seen when CQR parasites are energy-starved (Figs 1 and 2). Results are also presented for the CQS isolate GC03 (Fig. 5C). It is apparent that glucose starvation or collapsing the proton electrochemical gradient accelerates the efflux of CQ from the GC03 CQS strain to a rate close to that seen in the Dd2 CQR strain. Altogether, these results suggest that (i) the rapid efflux of CQ in CQR parasites is not directly coupled to the energy supply and (ii) the default (energy-deprived) state is one of rapid CQ efflux, even in CQS lines.

Data presented in Fig. 6 show that the rate of CQ efflux is controlled by mutations in PfCRT. Figure 6A shows data for the C2^GC03 line (GC03 transfected with its own wild-type pfcrt allele). This line is phenotypically CQS and displays a slow rate of CQ efflux that is accelerated upon energy deprivation. Figure 6B shows data for the C3^Dd2 line (GC03 transfected with the Dd2 CQR pfcrt allele). This line is phenotypically CQR and displays a rapid rate of CQ efflux that is independent of energy supply.

Fig. 6.

Influence of energy starvation on CQ efflux in lines harbouring wild-type and mutant alleles of pfcrt. CQ efflux was measured in the presence or absence of 10 mM glucose or 10 mM glucose with or without 10 μM FCCP.

A and B. Data for the C2^GC03 and C3^Dd2 lines respectively.

C and D. Data for the T76K-1^Dd2 and C-1^Dd2 recombinant clones respectively.

Data points are means of duplicate observations from at least two independent experiments.

Data presented in Fig. 6C and D show that reversing the K76T mutation (T76K-1^Dd2 – Dd2 T76K back mutant) is all that is required to change CQ efflux compared with the control CQR line C-1^Dd2 such that it resembles that of a CQS line (Fig. 5A). These data show that the amino acid change at codon 76 of the pfcrt gene plays a key role in determining the rate of CQ efflux, just as it does in the rest of the CQ phenotype (Lakshmanan et al., 2005).

Discussion

The mechanism of CQ accumulation – proton trapping, FP binding or both?

When attempting to explain the effects of uncoupling the proton electrochemical gradient it may be helpful to begin by considering two main theories to explain the parasite-specific accumulation of CQ. First, it has been proposed that CQ accumulates by a proton trapping mechanism. This drug is membrane permeable as a free base but is much less permeable in its monoprotonated and diprotonated forms (Ferrari and Cutler, 1991). It is predicted that protonated CQ will become trapped in acidic compartments such as the DV, potentially reaching millimolar concentrations (Homewood et al., 1972; Yayon et al., 1985). Second, it is known that CQ binds with high affinity to free FP, which is liberated upon digestion of haemoglobin in the parasite DV (Chou et al., 1980) and that this binding makes a significant contribution to the overall cellular uptake of CQ (Bray et al., 1999).

While it is clear that a mechanism of proton trapping will cause CQ to accumulate in the DV, it is interesting to note that CQ accumulation appears to be much higher in malaria parasites than it does in other organisms that also contain large acidic vesicles. For example, the uptake of CQ in CQS lines of P. falciparum (6000 fmol per 10⁶ cells per hour) is about 40-fold higher than seen with Dictyostelium discoideum (150 fmol per 10⁶ cells per hour) at the same extracellular CQ concentration of 100 nM (Saliba and Kirk, 1998; Naude et al., 2005). D. discoideum cells have a large acidic vacuole, as do P. falciparum; however, the former do not contain FP (Naude et al., 2005). Does CQ–FP binding account for the difference? It is notable that CQ accumulation by Entamoeba histolytica was shown to increase 20-fold when the cells were fed on erythrocytes rather than starch (Homewood et al., 1983). E. histolytica also contains a large acid vacuole and employ acid hydrolase enzymes to digest haemoglobin (and liberate FP) when fed on erythrocytes (Mora-Galindo et al., 2004). These data support the idea that binding of CQ to FP, in concert with proton trapping, are major factors in the specific accumulation of the drug by malaria parasites (Bray et al., 1999).

Here we show that for CQS lines, approximately half of the CQ accumulation is inhibited by the proton uncoupler FCCP or by glucose starvation over a 10 min period. It is tempting to speculate that the energy-insensitive element of CQ uptake reflects the binding of CQ to residual FP and that the energy-sensitive component reflects a combination of proton trapping and binding of CQ to newly liberated FP in the (now acidic) DV. It is interesting to note that CQ accumulation can be almost completely displaced by HF (Fig. 5A). HF is known to bind to FP with very high affinity (Egan et al., 1999) and so these results could reflect HF displacing CQ that is bound to FP in the DV. In addition, HF promotes the rapid efflux of CQ from CQS parasites, probably due to the alkalinization of the DV. HF is a monoprotic weak base (pKa = 8–9; Taillardat-Bertschinger et al., 2003) and is very lipophilic (Log D = 3.2). Therefore, it is likely that HF raises the DV pH by a proton shuttling mechanism (Ginsburg et al., 1989). This would allow CQ to leak from the DV of CQS parasites in its membrane permeable uncharged form. Neither VP nor PQ bind to FP (Bray et al., 1998) and as they do not alkalinize the DV, they do not significantly affect the efflux of CQ from CQS parasites (Fig. 5A). These data show that the movement of freely exchangeable CQ through membranes is extremely rapid and they are consistent with earlier reports suggesting that proton trapping and binding to FP are important factors in accounting for the specific uptake of CQ into P. falciparum-infected erythrocytes (Homewood et al., 1972; Chou et al., 1980; Yayon et al., 1985; Bray et al., 1998; 1999).

Is PfCRT an active CQ carrier?

At first glance, the proposal that mutant PfCRT is able to actively efflux CQ out of the DV is an attractive one (Sanchez et al., 2003; 2005; Naude et al., 2005). If so, our data would suggest that the energy source is in the form of the proton motive force across the DV membrane (Figs 1-3). However, in the default (de-energized) state, CQ accumulation is equal in CQS and CQR lines, reaching a level that is half way between that seen in energized CQS and CQR lines (Figs 1-3; Sanchez et al., 2003). It seems that despite hinging on just one critical PfCRT amino acid substitution (Lakshmanan et al., 2005), the CQR phenotype is quite complex, involving the loss of an energy-dependent uptake process as well as the acquisition of an energy-dependent efflux process (Krogstad et al., 1992; Sanchez et al., 2003). Thus, the efflux pump or active efflux carrier hypotheses do not appear to fully explain the CQ susceptibility phenotype. Furthermore, we present two key lines of evidence that CQ movement through PfCRT is not an active process.

First, we examine the energy dependence of CQ efflux. For some members of the DMT family, the energy-dependent efflux of substrates is directly coupled to the proton motive force (Jack et al., 2001). For example, the efflux of homoserine and threonine mediated by the RhtA transporter of Escherichia coli was shown to be completely inhibited by compounds that uncouple the proton gradient (Livshits et al., 2003). Here we have used a similar strategy to investigate the role of the proton gradient in PfCRT-mediated CQ transport. We have looked at the effect of proton uncouplers and glucose starvation on the rate of CQ efflux from parasite lines that were genetically modified with respect to PfCRT. The results of these experiments are presented in Fig. 5. We show that neither glucose starvation nor uncoupling the proton gradient has any effect on the rate of CQ efflux from CQR lines (Fig. 5B). In contrast, the resistance reversal agent VP was seen to dramatically reduce the rate of CQ efflux from CQR lines (Fig. 5B). These observations indicate that VP-inhibitable CQ transport via mutant PfCRT is not directly coupled to the energy supply. If mutant PfCRT were acting as an efflux pump or energy-dependent carrier then the efflux of CQ should have been slower when these cell lines were starved of energy, just as it is in the presence of VP. In CQS lines, we found that the rate of CQ efflux was actually increased to a rate approaching that seen in CQR lines, following glucose starvation or uncoupling of the proton gradient (Fig. 5C). Thus in the absence of any active transport process, CQS and CQR parasites both exhibit a rapid rate of CQ efflux (Fig. 5) as well as accumulating the same amount of CQ (Figs 1 and 2; Sanchez et al., 2003). Rather than supporting an active efflux mechanism, these observations suggest that efflux is not directly coupled to the energy supply.

Second, we will consider the trans-stimulation data reported by the Lanzer group. These investigators have shown that for CQR parasite lines harbouring mutant pfcrt alleles, the uptake of [³H]-CQ is first stimulated and then inhibited when the cells are pre-loaded with increasing concentrations of unlabelled CQ (Sanchez et al., 2003; 2004; 2005). For CQS parasite lines the uptake of [³H]-CQ is never stimulated but instead is progressively reduced under the same conditions (Sanchez et al., 2003; 2004; 2005). Although they could be suggestive of carrier-mediated CQ transport, these observations alone do not tell us whether CQ transport is in the inward or outward direction: stimulation of [³H]-CQ uptake could be due to acceleration of the transporter cycle by the outgoing unlabelled CQ (true trans-stimulation) or it could result from reduced efflux of [³H]-CQ due to competitive inhibition from the pre-loaded unlabelled CQ (simple cis-inhibition). The authors prefer the latter explanation and interpret these data in terms of a model in which CQR lines have an active efflux carrier. This carrier could be an ATP-dependent efflux pump (Sanchez et al., 2003) or more likely, a secondary active transporter (Sanchez et al., 2003; 2004; 2005). At low concentrations the pre-loaded unlabelled CQ is proposed to compete with [³H]-CQ that has entered the cell by passive diffusion for binding to the high-affinity carrier, blocking the active efflux of [³H]-CQ and hence increasing the net uptake of the label. The carrier is proposed to have a high affinity for CQ (K_p estimated to be in the low nanomolar range; Sanchez et al., 2003). Thus at higher pre-loading concentrations of unlabelled CQ the carrier is proposed to be saturated with unlabelled CQ resulting in a net reduction in the accumulation of [³H]-CQ. In CQS lines, the efflux carrier is not present and the pre-loading process results in a progressive decline in the accumulation of [³H]-CQ due to increasing competition for binding to FP (Sanchez et al., 2003; 2004; 2005).

In this model the [³H]-CQ and the unlabelled CQ are both on the same side of the membrane when they encounter the transporter, i.e. they are mixed together. If this were true, then there should be no need to pre-load the cells with unlabelled CQ. The same results should be obtained if the [³H]-CQ and the unlabelled CQ were mixed before the incubation. However, when we tested this directly, none of the CQR lines incubated with premixed CQ showed a significant rise in the [³H]-CQ accumulation ratio as the unlabelled CQ concentration was increased (Fig. 4B and C). The only way we can recreate the observations of the Lanzer group is if we pre-load with unlabelled CQ in the same way that they did (Fig. 4A). Thus the labelled and unlabelled CQ must initially be on opposite sides of the membrane to create the trans-stimulation effect. If this effect were indeed due to a carrier-mediated transport process, it follows that the carrier-mediated transport of unlabelled CQ would be in the outward direction while the carrier-mediated transport of the [³H]-CQ would be in the inward direction, i.e. mutant PfCRT would act as a bidirectional CQ carrier.

These observations appear incompatible with the concept of an efflux pump or an active efflux carrier. For an active transporter to pump the same substrate in the reverse direction there would normally be a requirement for the direction of the ion gradient driving the transporter to also be reversed. If PfCRT does indeed use the proton gradient (or any other gradient) to drive CQ transport, it is hard to understand how the direction of this gradient could have been reversed in these experiments. Our results suggest that mutant PfCRT is most unlikely to be acting as an active efflux pump or as a secondary active carrier. It is possible that mutant PfCRT may act as a passive bidirectional CQ carrier but this is not consistent with the observed lack of trans-stimulation in energy-starved CQR parasites (Sanchez et al., 2003).

Is PfCRT a voltage-gated channel?

An alternative explanation is that PfCRT might act as a channel or pore, with the mutant forms of the protein being permeable to CQ. The idea that the replacement of the positively charged lysine by the neutral threonine (K76T) in transmembrane domain 1 of PfCRT allows the leak of positively charged (protonated) CQ species out of the DV, is consistent with PfCRT acting as an aqueous channel (Warhurst et al., 2002; Johnson et al., 2004; Bray et al., 2005). The DV itself is acidified primarily by a V-type proton pumping ATPase (Saliba et al., 2003). In eukaryotic lysosomes these pumps are electrogenic, generating an inside-positive potential that is shunted in part by the inward movement of Cl⁻ ions. The ClC-3 chloride channel is thought to play a major role in this process (Nilius and Droogmans, 2003; Hara-Chikuma et al., 2005).

Warhurst et al. have suggested that the PfCRT protein sequence bears some resemblance to ClC channels of other organisms (Warhurst et al., 2002). Interestingly, it has been demonstrated that mutated pfcrt expressed in the yeast Pichia pastoris produces a chloride-dependent change in the acidity of inside-out membrane vesicles (Zhang et al., 2002). If PfCRT does indeed function as a lysosomal Cl channel, then we could explain the CQR phenotype, provided we assume that CQ transport is gated. The mechanism of ClC channel gating is complex and is not fully understood, but it is generally accepted that in the E. coli channel, opening is favoured by low external pH (periplasmic space) and positive voltage (Miller, 2006). It is important to note that the interior of eukaryotic intracellular vacuoles is effectively ‘outside’ the cell. If the analogy is correct this would suggest that under energized conditions the PfCRT channel would be open, allowing the efflux of charged CQ down the concentration gradient into the cytoplasm in CQR lines. If the proton pump is inhibited or uncoupled or if the energy supply is removed, the channel will be closed as the proton gradient is reduced and the depolarizing voltage is dissipated. When the channel is closed, CQS and CQR lines will behave in a similar way with respect to CQ: we would expect that removal of energy would reduce the accumulation of CQ in CQS lines and increase the rate of efflux as, by raising the lysosomal pH, it increase the concentration of unprotonated (permeable) CQ. In CQR lines CQ accumulation will increase because any residual charged CQ will be held inside the DV as the channel is closed (Figs 1 and 2).

The trans-stimulation results might also be explained in terms of a gated channel: in this case, the escape of high concentrations of pre-loaded unlabelled charged CQ in CQR lines would possibly cause a transient negative shift in the electrical potential inside the DV during the washing step and subsequent incubation with [³H]-CQ. This might be expected to close the putative channel, causing the transient stimulation of [³H]-CQ uptake that is observed. Sanchez et al. have suggested that their trans-stimulation data rule out the possibility of a leak mechanism (Sanchez et al., 2003; 2004; 2005). We respectfully disagree, and suggest that additional experiments/experimental approaches are required to distinguish between these two proposed mechanisms.

This gated channel hypothesis is compatible with the key role for the DV proton electrochemical gradient that is outlined here. It is also compatible with most (if not all) of the published data that were used earlier to support the active efflux hypothesis. Importantly it is also compatible with earlier observations that were difficult to reconcile with the concept of a CQ efflux pump or active carrier. These include the lack of stereospecificity of the CQR mechanism for isomers of CQ (Bray et al., 1996), the lack of stimulation of CQ uptake at reduced temperature (Sanchez et al., 1997), the importance of physicochemical properties of 4-aminoquinolines to their recognition by the CQ resistance mechanism (Bray et al., 1996) and selective stimulation of CQ uptake by protonophoric compounds in CQR isolates (Bray et al., 1999).

In conclusion we present some surprising findings that stand against the hypothesis that PfCRT-mediated CQR is mediated by an efflux pump or secondary active carrier. It is possible that PfCRT could act as a passive CQ exchanger in CQR lines but our observations can perhaps be more readily explained in terms of a gated CQ leak mechanism.

Experimental procedures

Parasite lines, culture and chemicals

Dd2 is a CQR P. falciparum line from Indochina. GC03 is a CQS progeny of the Dd2 × HB3 genetic cross that led to the identification of pfcrt (Wellems et al., 1990). C2^GC03 and C3^Dd2 were produced by allelic modification of the pfcrt locus in the GC03 line and express the GC03 wild-type CQS and the Dd2 mutant CQR pfcrt alleles respectively (Sidhu et al., 2002). C-1^Dd2 and T76K-1^Dd2 are also recombinant clones produced by allelic modification of the pfcrt locus in Dd2 parasites, with C-1^Dd2 expressing the parental Dd2 allele whereas T76K-1^Dd2 expresses a variant Dd2 ‘back-mutant’ allele in which the mutant threonine codon 76 (T76) was replaced with the wild-type lysine codon 76 (K76) (Lakshmanan et al., 2005). P. falciparum parasite lines were maintained in continuous in vitro culture using standard methods (Trager and Jensen, 1976). Briefly, cultures contained a 2% suspension of O⁺ erythrocytes in RPMI 1640 (R8758) medium supplemented with 10% pooled human AB⁺ serum, 25 mM HEPES (pH 7.4) and 20 μM gentamicin sulphate. All CQ transport experiments were performed on highly synchronous trophozoite-stage cultures, 24 h post invasion. Some experiments employed magnet-purified trophozoites (Sanchez et al., 2003). Chemicals were all purchased from Sigma.

Chloroquine transport experiments (general)

Sanchez et al. (2003) have suggested that magnet-purified preparations of trophozoite-infected erythrocytes at high parasitaemia are superior to synchronized trophozoite cultures of lower parasitaemia for use in drug transport experiments. We acknowledge that this technique offers the advantage of reducing the non-specific components of drug transport by effectively eliminating the uninfected erythrocyte compartment. However, we found that a large proportion of parasites (up to 40% in some cases) can become liberated from their host cells during the subsequent experiments. Nonetheless, we have employed magnet-purified trophozoites in representative experiments in every case. These results were largely similar to those obtained with synchronized cultures; however, they have not been included because of concerns regarding parasite viability and inoculum effect.

For experiments performed in the absence of glucose, parasites were first washed at room temperature in RPMI without glucose, followed by a 15 min incubation in RPMI without glucose in a 37°C water bath, followed by a further 37°C wash in RPMI without glucose, prior to starting the experiment. ATP concentrations were tested as described (Sanchez et al., 2003), and cells were found to be completely depleted of ATP.

Triplicate samples of synchronized trophozoite stage P. falciparum cultures (approximately 5% parasitaemia) or magnet-purified trophozoites (approximately 97% parasitaemia) were suspended in RPMI containing 10 mM HEPES, pH 7.4, with or without 10 mM glucose. In the case of synchronized trophozoite cultures the inoculum size was adjusted to 2. The inoculum size is a parameter defined (without units) as the product of parasitaemia and haematocrit. Using an inoculum size of 2 ensures that the depletion of [³H]-CQ from the medium is always less than 10% (not shown). In the case of the magnet-purified trophozoites the inoculum size was identical to that used by Sanchez et al. (2003). Suspensions of cells were placed in a 37°C water bath and after sufficient time for the temperature to equilibrate, 1–10 nM [³H]-CQ was added to initiate the experiments. Following incubation, aliquots of the suspension were overlaid onto Dow Corning 550 silicon oil in 1.5 ml Eppendorf tubes. The tubes were then centrifuged and the bottom of the tube containing the cell pellet was cut off. Cell pellets were lysed and removed from the tube bottom by the addition of 100 ml distilled water and then solubilized and decolourized by the addition of a 100 ml cocktail containing five parts tissue solubilizer, two parts H₂O₂ (30%) and two parts glacial acetic acid. Samples were then counted by liquid scintillation counting. In some cases (detailed in the figure legends) unlabelled CQ was also added with the [³H]-CQ to initiate the experiments. For most experiments, the medium did not contain added bicarbonate. Using bicarbonate-free medium (as supplied) in atmospheric air at pH 7.4 will result in a bicarbonate concentration at equilibrium of approximately 100–200 μM (S. Hladky, pers. comm.). In some cases (detailed below) 23 mM sodium bicarbonate was added to the medium. In all experiments, the inoculum was adjusted to ensure that CQ depletion from the medium was less than 10% at the end of the experiment. In experiments employing synchronized trophozoite cultures rather than magnet-purified trophozoites, the counts corresponding to the CQ uptake of a similar number of uninfected erythrocytes were subtracted from the total. When appropriate, the saturable component of CQ uptake was calculated by subtracting from the average total counts (minus the small amount of CQ taken up by uninfected erythrocytes) those resulting from non-saturable CQ uptake into infected erythrocytes (calculated using 100 μM external CQ; Bray et al. 1998).

A number of transport experiments were performed in the presence of 10 μM FCCP, 5 μM VP or 200 nM HF. These compounds were first dissolved in DMSO at a stock concentration of 10 mM. Control samples were treated with appropriate concentrations of DMSO alone. FCCP was not found to affect the levels of ATP in the parasite under the experimental conditions used (data not shown).

Chloroquine efflux experiments

Krogstad et al. (1987) have reported that CQ efflux is rapid from CQR isolates (t_1/2 ~2 min) but much slower from CQS parasites (t_1/2 ~20 min). On the contrary, we reported that CQ efflux is similarly rapid (t_1/2 2–5 min) regardless of whether the parasites are CQS or CQR (Bray et al., 1992). We have found that this discrepancy is due to two critical differences in the protocol employed in the two studies. First, Krogstad et al. washed the pre-loaded parasite pellet in medium without [³H]-CQ before measuring zero-trans efflux of radiolabel. In contrast, in the Bray et al. (1992) study the pellet was used unwashed. Second, Krogstad et al. used medium without added bicarbonate whereas Bray et al. used normal culture medium containing 23 mM bicarbonate. Without washing the pellet and using medium containing bicarbonate we did not observe any major differences in the initial CQ efflux rates as reported earlier (Bray et al., 1992). However, we were able to reproduce the observations of Krogstad et al. by using their protocol, i.e. by washing the pellet and using medium without added bicarbonate (Krogstad et al., 1987).

For the efflux experiments, cells were pre-loaded for a 15 min period by suspending them in medium without added bicarbonate containing various concentrations of [³H]-CQ in a 37°C water bath. The concentration of [³H]-CQ used in the pre-loading procedure was adjusted during preliminary experiments so that the CQ concentration in the cell pellets at the start of the efflux experiment was similar for each cell line and experimental condition. In experiments that examined the effect of glucose deprivation on the efflux rate, the cells were glucose deprived during the pre-loading procedure. Similarly, to measure the effect of FCCP, this compound was added at the beginning of the pre-loading procedure. In contrast, the effect of HF, PQ and VP was determined by adding these compounds at the start of the efflux experiment rather than during the pre-loading procedure. The range of [³H]-CQ concentrations used was between 1.5 and 35 nM. After pre-loading, cells were collected by centrifugation and washed once in medium without bicarbonate at room temperature. Then cells were again collected by centrifugation and the efflux experiment was initiated by suspending the cell pellet in medium without added bicarbonate (pre-warmed to 37°C in a water bath). The cell suspension was placed in the water bath and samples were removed and the cells rapidly centrifuged through silicon oil at the required time points. The cell pellets and supernatants were processed and analysed as described above.

Trans-stimulation experiments

These experiments were performed as described in Sanchez et al. (2003) with the exception that suspensions of synchronized trophozoite cultures were used instead of magnet-purified trophozoite suspensions.

Measurement of CQ intracellular compartmentalization

Plasmodium falciparum-infected erythrocytes were loaded with 2.5 nM [³H]-CQ for 10 min at 37°C in bicarbonate-free RPMI with or without inhibitors or chemosensitizers. Purification of freed parasites and subsequent isolation of DV were performed using saponin lysis and hypo-osmotic lysis (Saliba et al., 1998; Biagini et al., 2003) techniques respectively. Preliminary experiments in which the parasite cytoplasm and DV were loaded with different impermeable fluorescent probes showed that the cytoplasmic compartment (but not the DV) was completely disrupted during the isolation of DV. All manipulations were carried out at 0°C to prevent CQ movement. Prior to the isolation of DV, a sample of free parasites was taken for scintillation counting. Free parasite and isolated DV fractions were processed for liquid scintillation counting as described above.